基于PSO-BP的超声滚挤压轴承套圈表面加工硬化程度预测  

Prediction on degree of work hardening for surface of bearing ring by ultrasonic rolling extrusion based on PSO-BP

作　　者：朱其萍 徐红玉[1,2] 王晓强 刘志飞[1,2] 刘东亚 Zhu Qiping;Xu Hongyu;Wang Xiaoqiang;Liu Zhifei;Liu Dongya(College of Mechanical and Electrical Engineering,Henan University of Science and Technology,Luoyang 471003,China;Collaborative Innovation Center of Advanced Manufacturing of Mechanical Equipment,Luoyang 471003,China)

机构地区：[1]河南科技大学机电工程学院,河南洛阳471003 [2]机械装备先进制造河南省协同创新中心,河南洛阳471003

出　　处：《锻压技术》2021年第11期190-196,共7页Forging & Stamping Technology

基　　金：国家自然科学基金资助项目(U1804145);国家重点研究专项(2018YFB2000405)。

摘　　要：为了提高轴承套圈表面质量、延长轴承使用寿命,对超声滚挤压加工参数与轴承套圈表面加工硬化程度之间的影响规律进行分析。提出采用PSO-BP神经网络模型进行预测,建立以加工过程中4个主要参数为输入、加工硬化程度为输出的神经网络模型,采用粒子群优化算法对BP神经网络模型的权值和阈值进行优化,并对该模型进行了验证。结果表明:采用PSO算法优化的BP神经网络模型可有效地避免网络陷入局部最优的问题,具有更好的泛化能力,预测精度高,预测相对误差在0.5%以内,预测平均绝对百分比误差降低了0.378%。In order to improve the surface quality of bearing ring and prolong the service life of bearing, the influence laws of ultrasonic rolling extrusion parameters on the degree of working hardening for the surface of bearing ring were analyzed, and PSO-BP neural network model was proposed to make prediction to establish a neural network model that takes four main parameters in the processing as input and the degree of work hardening as output. Then, weights and thresholds of BP neural network model were optimized by the particle swarm optimization algorithm, and the model was verified. The results show that the BP neural network model optimized by PSO algorithm can effectively avoid the network falling into the local optimal problem, which has better generalization ability and high prediction accuracy. It is shown that the relative error of the prediction is within 0.5%, and the average absolute percentage error of the prediction is reduced by 0.378%.

关 键 词：超声滚挤压 轴承套圈 加工硬化 BP神经网络 粒子群算法 

分 类 号：TG376.1[金属学及工艺—金属压力加工]